Interpretation of estimates

Monitoring prevalence (the number of people with diabetes at any one time divided by total population) and incidence (new cases of diabetes over a period of time divided by the total population from which cases arise) are important indices of disease burden and useful for monitoring the impact of preventive interventions.

In each edition of the IDF Diabetes Atlas, we estimate diabetes prevalence based on the best quality data available at the time of analysis as judged by the international diabetes experts who comprise the IDF Diabetes Atlas Committee. It is important to highlight that the diabetes estimates presented in this edition have varying degrees of uncertainty. Thus, these estimates must be interpreted along with their confidence intervals (CI) where the true estimate may lie anywhere between the upper and lower bounds. A wide CI indicates an imprecise estimate while a narrow interval reflects more precision.

Depending on study design, sample size, type of measurements performed, definitions and analysis methods, these estimates can vary markedly between studies and between countries as well as over time. Therefore, changes in the magnitude of prevalence of individual countries from edition to edition and comparisons between countries should be treated with caution. Additionally, predicted estimates for the future are based only on projected changes in age, sex and rural-urban residence location as defined by the UN.

Gathering and selecting data sources

The data used for the estimation of diabetes prevalence in this edition of the IDF Diabetes Atlas were obtained from a variety of sources. The vast majority were extracted from peer-reviewed publications and national health surveys including selected WHO STEPwise approach to surveillance (WHO STEPS) studies.¹ Data from other official sources such as registries and reports from health regulatory bodies were also used, provided there was sufficient information to assess their quality. Data sources with sufficient methodological information on key areas of interest such as method of diagnosis and representativeness of the sample were also included. Given the importance of age as a major determinant for diabetes prevalence, only studies with at least three age-specific estimates were included. Data sources published before 2005 were excluded, except when a country lacked a more recent study.

In total, 41 WHO STEPS were included as data sources, of which 13 were included for the 10^th edition of the IDF Diabetes Atlas for the first time. WHO STEPS studies that have been recently shown to overestimate diabetes prevalence² were excluded from this edition of the IDF Diabetes Atlas. Furthermore, territories that had a population less than 50,000 were excluded. As a result, this edition presents data for 215 countries and territories compared to the previous edition, which had 211. The territories added in this edition are American Samoa, Holy See, Isle of Man, Mayotte and the Northern Mariana Islands.

In addition, data sources published between January 2019 and December 2020 were screened and added to the existing database if they met the inclusion criteria mentioned below. This added 81 data sources from 67 countries to the existing database (Map 2.1).

Map 2.1Countries and territories with in-country data sources on diabetes

To evaluate the quality of available data, each data source was scored, as in previous IDF Diabetes Atlas editions, using an analytical hierarchy process (AHP)³ taking into account the criteria mentioned in Figure 2.1. In this table, the classification possibilities for each of the criteria are presented, arranged from highest to the lowest degree of preference. In total, 219 out of 860 included data sources (25.5%) published since 2005 met the rigorous inclusion criteria or were included by expert consensus for this 10^th edition.

The final score of a data source is the summary of all scores on the five criteria mentioned in Figure 2.1. Data sources that received a score over a certain threshold (agreed in consensus with members of the IDF Diabetes Atlas Committee) were used to generate the estimates and projections. Preference was given to data sources that were nationally representative, conducted in the past five years, published in peer-reviewed journals and were based on the objective measurement of diabetes status (rather than self-reported).^†

Figure 2.1

Classification of diabetes data sources.

Estimating diabetes prevalence and projections for the future

After the selection of data sources, the reported age- and sex-specific data in each data source were smoothed using a logistic regression model. If more than one data source was available for an individual country, the country level diabetes estimates were derived using an average of prevalence estimated from various data sources, with each weighted by the quality score based on the AHP scoring. This permitted the higher quality studies to contribute more to the final country estimate. The details of the logistic regression model are described in a previous publication⁴ and any changes and development of the methods are summarised more recently by Sun et al.⁵

For each country, the age- sex- and urban/rural-specific diabetes estimates were generated. For studies that did not present results stratified by urban and rural area, the estimated ratio of urban to rural prevalence of disease from the original data or United Nations Population Division (UNPD)⁶ approximation was used to distribute the results by urban and rural prevalence. Prevalence estimates were then aggregated to produce estimates for the seven IDF Regions and countries in the four World Bank income classification categories.

The 2021 population data from the UNPD were used in estimating the number of people with diabetes for each nation. In order to project diabetes estimates forward to the years 2030 and 2045, population projections for 2030 and 2045 from the UNPD for each nation were used. The 2030 and 2045 diabetes projections assume that diabetes prevalence does not change for each age group, but take into account the changes in population age structure and degrees of urbanisation.⁷ This approach is likely to underestimate future diabetes prevalence as it does not take into account changes in prevalence of obesity and other risk factors that might result in a higher diabetes incidence.

However, estimating diabetes projections for 2030 and 2045 in this way allows comparison with projections made, for the same years, in previous editions of the IDF Diabetes Atlas. Increases or decreases in diabetes prevalence in specific countries in this edition compared to the previous editions of the IDF Diabetes Atlas may be the result of updates or changes in data sources and may not be a complete or precise reflection of actual changes in diabetes prevalence occurring in that country.

Extrapolating data

One third of countries or territories with more than 50,000 habitants (71 countries out of 215 countries or territories, 33%) do not have in-country data sources on diabetes prevalence that fulfil the IDF Diabetes Atlas inclusion criteria. Under such circumstances, estimates were generated by extrapolation using diabetes prevalence data from countries that are similar in terms of ethnicity⁸, language⁹, World Bank income classification¹⁰ and geographic location.

Extrapolated estimates are less reliable than estimates based on national data sources and should therefore be interpreted with caution. Countries with extrapolated estimates are designated in the country summary table (Appendices) and Map 2.1. The necessity of extrapolation emphasises the importance of conducting high quality studies worldwide that help to address gaps in diabetes prevalence information.

Estimating confidence intervals

Confidence intervals are provided to indicate the degree of uncertainty around each of the estimates. In order to calculate these, two separate analyses were performed: a jack-knife analysis of the sensitivity of the global prevalence estimate to the study selection process and a simulation study to access raw data uncertainty.

The confidence interval for each age group, sex and country was constructed based on combining the minimum and maximum for simulation analyses and the study selection jackknife analyses. These procedures are described more fully elsewhere.⁵

Standardisation of estimates

It is important to appreciate that the IDF Diabetes Atlas data is standardised to two different populations and both types of standardised estimates are referred to in the forthcoming chapters.

The total numbers of persons with diabetes, IGT and IFG for each country were calculated by applying the calculated age-, sex- and urbanisation- specific prevalence rates to that country’s 2021 age-, sex- and urban/rural setting distribution as estimated by the United Nations Population Prospects 2019 Revision.^6,11

This provides a prevalence estimate standardised to the national population, as indicated in the table and figure footnotes: ‘standardised to each national population’. This method was used to produce individual country standardised diabetes prevalence estimates. However, to permit comparison of diabetes prevalence between countries, additional ‘comparative’ estimates were also calculated.

These were produced by standardising 2021 prevalence estimates from each country to the age structure of an estimated UN world population. They are referred to as ‘comparative prevalence’ and indicated with a footnote: ‘standardised to world population’. This latter standardisation approach removes the effect of differences in the age structure between countries. The comparative diabetes prevalence in 2030 and 2045 was calculated using the UN projected global age structures for 2030 and 2045, respectively.⁶

Estimating undiagnosed diabetes prevalence

The early detection of diabetes and initiation of treatment is extremely important in the management of diabetes and prevention of complications. The longer a person has diabetes but remains undiagnosed, the greater the risk of developing complications. People are defined as having undiagnosed diabetes when their blood glucose levels would satisfy the diagnostic criteria for diabetes, but the diagnosis has not been confirmed by a doctor.

Population-based scientific studies allow us to estimate the prevalence of undiagnosed diabetes worldwide. A sample of the population is surveyed to assess how many people have diabetes. Those who say they do not have diabetes are then tested. This helps establish the total prevalence of people already diagnosed with diabetes, and those who tested positive for diabetes in this population sample. The number of undiagnosed people as a proportion of the total number of people living with diabetes is then extrapolated to calculate country-level estimates for undiagnosed diabetes. The proportion of undiagnosed diabetes may differ greatly across countries with different access to healthcare, with less access likely related to a higher proportion with this undiagnosed condition.

It is important to keep in mind that diagnostic tests are based on different biochemical processes and often yield different results.¹² Common tests, such as an oral glucose tolerance test (OGTT), fasting blood glucose (FBG) or a haemoglobin A1c (HbA1c) test detect different subgroups of people with diabetes and those groups only partially overlap. While this means studies may report different proportions of undiagnosed people, any diagnostic method for diabetes even if it is not able to detect all cases, can be useful in detecting diabetes earlier.

For this edition of the IDF Diabetes Atlas, we have assembled all published studies reporting undiagnosed diabetes that met defined selection criteria, regardless of the way diabetes was detected. The average of the estimates was calculated for countries that reported data on estimates of undiagnosed diabetes. However, in countries without in-country data sources, the undiagnosed proportion was approximated by the average of the estimates from countries with data sources within the same IDF Regions and World Bank Income Group (Map 2.2). Further details of this methodology can be found in previous IDF publications.¹³

Map 2.2Countries and territories with data sources on the proportion of adults (20–79 years) with previously undiagnosed diabetes

Estimating the incidence and prevalence of type 1 diabetes in children and adolescents

The incidence and prevalence estimates of type 1 diabetes in children and adolescents (0–14 and 0–19 years of age) were produced by the 10^th Edition of the IDF Diabetes Atlas Type 1 Diabetes in Children and Adolescents Special Interest Group, using methodology from the 9th Edition of the IDF Diabetes Atlas as previously described¹⁴, along with published prevalence data, when available.

The scientific literature was searched, without language restrictions, for data sources that contained population-based studies on the incidence of type 1 diabetes (new cases each year) or prevalence (existing total cases) in children and adolescents aged up to 20 years. If more than one study was available for a country, the following criteria were applied to select the most suitable study: recent, population-based studies, high (≥90%) ascertainment level, covering a large part of the country, providing age- and sex-specific rates, and including the age ranges 0–14 and 15–19 years. For some countries where two or more studies met these criteria to an equal extent, results were combined by averaging age- and sex-specific rates.

If a country did not have any information available, the incidence rate for ages under 15 years was estimated using data from a similar country, based on geographical proximity, income and ethnicity. For ages 15–19 years, the incidence rate was estimated using the average regional ratio of incidence rates in the 0–14 and 15–19 years age groups.

Prevalence estimates were then derived from these incidence rates, and both were applied to UN population estimates for respective countries to obtain estimates of the numbers of incident and prevalent cases. However, there was a need to adjust prevalence estimates derived from the incidence rates to allow for case fatality, particularly in low-income countries. A mortality-adjusted prevalence was calculated for each country, based on a standardised mortality ratio for people with type 1 diabetes predicted from the country’s infant mortality rate (IMR) using a relationship derived in a systematic review of mortality studies in children with type 1 diabetes.^14,15 IMR data were obtained from the WHO Global Health Observatory data repository.¹⁶ For countries not included in the repository, the Central Intelligence Agency World Factbook¹⁷, UN country profile¹⁸ or IndexMundi¹⁹ were used.

A search was also made for type 1 diabetes prevalence studies in children and adolescents. Studies were required to have no data older than the year 2000, have sound, clearly-defined methodology where prevalence was measured rather than imputed from incidence, be country-wide or from a representative part of a country, have ascertainment estimated at ≥90%, and be within five years of the dates of an incidence study.

Publications from 12 countries were found and these published prevalence results were used in place of the calculated prevalence method for these countries.

Estimating the incidence and prevalence of youth-onset type 2 diabetes

Youth-onset type 2 diabetes, broadly defined as type 2 diabetes when diagnosed under 20 years of age, is increasingly recognised as an emerging chronic disease in children and adolescents. However, nationally representative epidemiologic data to monitor the occurrence of youth-onset type 2 diabetes are lacking in many regions – most notably in sub-Saharan Africa.

There are many challenges to collecting good quality data on youth-onset type 2 diabetes, in particular the categorisation of diabetes types, given the overlap in clinical presentations of type 2 diabetes, type 1 diabetes and monogenic diabetes.

The comparison of youth-onset type 2 diabetes incidence and prevalence within and across countries and by time period is difficult due to differences in methods of case ascertainment and completeness, variation of age ranges reported as youth-onset, and varying quality of information. The proportion of youth with undiagnosed type 2 diabetes across different regions may also impact the overall epidemiology of youth-onset type 2 diabetes.

Estimating the prevalence of impaired glucose tolerance and impaired fasting glucose

Data sources for impaired glucose tolerance (IGT) and impaired fasting glucose (IFG) prevalence were identified and selected according to criteria previously described (See Chapter 1). The urban and rural IGT and IFG prevalence ratios were calculated according to the weighted average of the ratios reported in various data sources from seven IDF Regions and the World Bank country classifications by income.

A logistic regression model was used to estimate the prevalence of IGT and IFG by country. The number of studies that satisfied the selection criteria was limited to 57 studies (from 48 countries) for IGT and to 49 studies (from 44 countries) for IFG. The prevalence estimates for the remaining countries were extrapolated from countries deemed to be similar, as for total diabetes prevalence (Map 2.3).

Map 2.3Countries and territories with data sources on impaired glucose tolerance in adults (20–79 years)

Estimating the prevalence of hyperglycaemia in pregnancy

Data sources reporting age-specific prevalence of gestational diabetes mellitus (GDM) and diabetes first detected in pregnancy were searched²⁰ and selected according to the criteria described previously.²¹ UN fertility projections²² and IDF estimates of diabetes were used to calculate the total percentage of live births affected by hyperglycaemia in pregnancy (HIP).

All studies were scored according to the diagnostic criteria used, the year the study was carried out, study design, the representativeness of the sample and the screening approach. Studies which met our pre-defined threshold were then selected to calculate country-level estimates. For this edition of the IDF Diabetes Atlas, 58 studies from 47 countries were used to estimate country-level, age-specific prevalence of HIP using a generalised linear regression model (Map 2.4). The detailed methods for estimation of prevalence of HIP have been described previously.²¹

Map 2.4Countries and territories with selected data sources on hyperglycaemia in pregnancy (adults 20–49 years), 2021

It should be noted that the method for selecting data sources was updated in the 9th edition of the IDF Diabetes Atlas. Thus, any comparison of the prevalence estimates from the 9^th and 10^th editions with those of previous editions must be viewed with caution. The changes in the selection of data sources include:

• ■ International Association of Diabetes and Pregnancy Study Group (IADPSG) diagnostic criteria have been given more weight in this edition compared to previous editions.
• ■ A new criterion, termed “screening approach”, has been added that includes the following options: universal one step, selective, two or more steps, and selective two or more steps.

Map 2.5Countries and areas with data sources on impaired fasting glucose in adults (20–79 years), 2021

Estimating diabetes-related mortality

The total number of deaths attributable to diabetes by country was calculated by combining information on the number of annual deaths from all-causes stratified by age and sex²³, age- and sex-specific mortality relative risks in people with diabetes compared to those without diabetes, and country-specific diabetes prevalence by age and sex for the year 2021. Relative risks attributable to diabetes are derived from cohort studies comparing death rates in those with and without diabetes.^24,25 This method of estimating diabetes-related mortality is described in more detail elsewhere.^26–28

Estimating the economic impact of diabetes

The direct cost estimates in this edition of the IDF Diabetes Atlas were calculated using an attributable fraction method, which relies on the following inputs:

• ■ IDF Diabetes Atlas estimates of diagnosed and undiagnosed diabetes prevalence for each country and for each age and sex sub-group, stratified by rural and urban setting.
• ■ UN population estimates for 2021 and UN population projections for 2030 and 2045.
• ■ WHO global health expenditures per capita for 2018 (latest available data).
• ■ The ratios of health expenditures for people with diabetes compared to people without diabetes, stratified by age, sex, rural versus urban setting, diagnosed and undiagnosed diabetes, and income per IDF Region.

The WHO definition of health expenditure includes provision of health services (preventive and curative), family planning activities, nutrition activities, and emergency aid designated for health, but does not include provision of water and sanitation services. The definition includes health expenditures from both public and private sources.²⁹ The same method was used as in the previous editions to distribute the total health expenditure in a given country into expenditure by age and sex.³⁰

Another critical component of the analyses is the ratio of health expenditures for people with diabetes (diagnosed or undiagnosed) compared to those without diabetes. Since the publication of the IDF Diabetes Atlas 8th edition, these ratios have been refined by the work of Bommer et al. (2017)³¹, providing estimates for this ratio with much more specificity in relation to age, sex, rural versus urban setting, whether diabetes is diagnosed, region, and income levels of countries.

The diabetes-related health expenditure estimates are presented in US dollars (USD), and in international dollars (ID), as well as a percentage of total health expenditures and of gross domestic product (GDP). IDs account for local purchasing power and facilitate direct cross-country comparisons of health expenditures. Health expenditures for diabetes as a percentage of total health expenditures and of GDP reflect the direct economic burden of diabetes to a national economy.

References

1.

World Health Organization. STEPS: A framework for surveillance. Geneva; 2003

2.

Lin S, Rocha VM, Taylor R. Artefactual inflation of type 2 diabetes prevalence in WHO STEP surveys. Trop Med Int Health;2019 Feb 1 [PubMed: 30706604]

3.

Saaty TL. Relative measurement and its generalization in decision making why pairwise comparisons are central in mathematics for the measurement of intangible factors the analytic hierarchy/network process. Revista de la Real Academia de Ciencias Exactas, Fisicas y Naturales Serie A Matematicas. 2008 Sep;102(2):251–318

4.

Guariguata L, Whiting D, Weil C, Unwin N. The International Diabetes Federation diabetes atlas methodology for estimating global and national prevalence of diabetes in adults. Diabetes Res Clin Pract. 2011 Dec;94(3):322–32 [PubMed: 22100977]

5.

Sun H, Saeedi P, Karuranga S, Pinkepank M, Ogurtsova K, Duncan BB, et al. IDF Diabetes Atlas: Global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045. Diab Res Clin Pract. 2021 (in press) [PubMed: 34879977]

6.

United Nations. World Population Prospects: The 2017 Revision. New York; 2017

7.

United Nations. 2018 Revision of World Urbanization Prospects. New York; 2018

8.

Central Intelligence Agency. The World Factbook, Ethnic groups. Washington, DC; 2015

9.

Central Intelligence Agency. The World Factbook, Languages. Washington, DC; Languages; 2015

10.

The World Bank. World Bank Country and Lending Groups; 2015.

11.

Ahmad OB, Boschi-Pinto C, Lopez AD, Murray CJ, Lozano R, Inoue M. Age Standardization of Rates: A new WHO Standard.:14

12.

Galaviz KI, Narayan KMV, Lobelo F, Weber MB. Lifestyle and the Prevention of Type 2 Diabetes: A Status Report. Am J Lifestyle Med. 2015;12(1):4–20. Published 2015 Nov 24. doi:10.1177/1559827615619159 [PMC free article: PMC6125024] [PubMed: 30202378] [CrossRef]

13.

Beagley J, Guariguata L, Weil C, Motala AA. Global estimates of undiagnosed diabetes in adults. Diabetes Res Clin Pract 2014;103:150–60. https://doi​.org/10.1016/j​.diabres.2013.11.001. [PubMed: 24300018]

14.

Patterson CC, Karuranga S, Salpea P, Saeedi P, Dahlquist G, Soltesz G, Ogle GD. Worldwide estimates of incidence, prevalence and mortality of Type 1 Diabetes in children and adolescents: Results from the International Diabetes Federation Diabetes Atlas. Diabetes Res Clin Pract 2019; 157:107842 [PubMed: 31518658]

15.

Morgan E, Cardwell CR, Black CJ, McCance DR, Patterson CC. Excess mortality in Type 1 diabetes diagnosed in childhood and adolescence: a systematic review of population-based cohorts. Acta Diabetol. 2015 Aug;52(4):801–7 [PubMed: 25585594]

16.

World Health Organization. Global Health Observatory data repository, Life tables by country WHO. [Internet]. 2019. Available from: http://apps​.who.int/gho/data/node​.main.LIFECOUNTRY?lang=en

17.

Central Intelligence Agency (CIA). The World Factbook [Internet]. Available from: https://www​.cia.gov/the-world-factbook/

18.

United Nations. UNdata Country Profile. New York 2019. Available from: http://data​.un.org/en/index.html

19.

IndexMundi – Country facts (2019). Available from: http://www​.indexmundi.com

20.

Yuen L, Saeedi P, Riaz M, et al. Projections of the prevalence of hyperglycaemia in pregnancy in 2019 and beyond: Results from the International Diabetes Federation Diabetes Atlas, 9th edition. Diabetes Res Clin Pract. 2019 Nov;157:107841. DOI: 10.1016/j.diabres.2019.107841. Epub 2019 Sep 10 [PubMed: 31518656] [CrossRef]

21.

Linnenkamp U, Guariguata L, Beagley J, et al. The IDF Diabetes Atlas methodology for estimating global prevalence of hyperglycaemia in pregnancy. Diabetes Res Clin Pract. 2014 Feb;103(2):186–96. DOI:10.1016/j.diabres.2013.11.004 [PubMed: 24300016] [CrossRef]

22.

United Nations. World population prospects (2017 revision). New York; 2017

23.

World Health Organization. Global Health Estimates 2019 Summary Tables [Internet]. Geneva; 2019 [cited 2021 Jun 21]. Available from: https://www​.who.int/data​/gho/data/themes​/mortality-and-global-health-estimates​/ghe-leading-causes-of-death

24.

Bracco PA, Gregg EW, Rolka DB, Schmidt MI, Barreto SM, Lotufo PA, et al. A nationwide analysis of the excess death attributable to diabetes in Brazil. J Glob Health. 2020 Jun;10(1):010401. doi: 10.7189/jogh.10.010401. PMID: 32257151; PMCID: PMC7101024 [PMC free article: PMC7101024] [PubMed: 32257151] [CrossRef]

25.

Bragg F, Holmes MV, Iona A, Guo Y, Du H, Chen Y, et al. Association Between Diabetes and Cause-Specific Mortality in Rural and Urban Areas of China. JAMA. 2017 Jan 17;317(3):280–289. doi: 10.1001/jama.2016.19720. PMID: 28114552; PMCID: PMC6520233 [PMC free article: PMC6520233] [PubMed: 28114552] [CrossRef]

26.

Saeedi P, Salpea P, Karuranga S Petersohn I, Malanda B, Gregg EW, et al. Mortality attributable to diabetes in 20–79 years old adults, 2019 estimates: Results from the International Diabetes Federation Diabetes Atlas, 9th edition. Diabetes Res Clin Pract. 2020 Apr;162:108086. doi: 10.1016/j.diabres.2020.108086. Epub 2020 Feb 15 [PubMed: 32068099] [CrossRef]

27.

Roglic G, Unwin N. Mortality attributable to diabetes: estimates for the year 2010. Diabetes Res Clin Pract. 2010 Jan;87(1):15–9 [PubMed: 19914728]

28.

IDF Diabetes Atlas Group. Update of mortality attributable to diabetes for the IDF Diabetes Atlas: Estimates for the year 2011. Diabetes Research and Clinical Practice. 2013 May;100(2):277–9 [PubMed: 23506763]

29.

World Health Organization. Global Health Expenditure database. [Internet]. 2021. Available from: https://apps​.who.int/nha/database

30.

Williams R, Karuranga S, Malanda B, et al. Global and regional estimates and projections of diabetes-related health expenditure: Results from the International Diabetes Federation Diabetes Atlas, 9th edition. Diabetes Res Clin Pract. 2020;162:108072. [PubMed: 32061820]

31.

Bommer C, Heesemann E, Sagalova V, Manne-Goehler J, Atun R, Bärnighausen T, et al. The global economic burden of diabetes in adults aged 20–79 years: a cost-of-illness study. Lancet Diabetes Endocrinol. 2017;5(6):423–30 [PubMed: 28456416]